Language_Model
-
BIRD-RL: Scaling Agentic Reinforcement Learning over Stateful Data-Centric Environments
2026-10-21 08:50:39Ge Qu, Jinyang Li, Xiaohan Xu, Xiaolong Li, Bowen Qin, Nan Huo, Ziwei Tang, Chenhao Ma, Reynold ChengAbstract
While agentic reinforcement learning (RL) has emerged as a promising paradigm for training small language models (SLMs), existing frameworks are predominantly tool-centric and struggle with data-centric environments. Specifically, applying tool-centric frameworks to stateful tasks like SQL debugging introduces severe spatial inefficiencies since proper environment isolation typically requires a separate database container per rollout. Naively reusing databases without tracking evolving states across multi-turn interactions causes trajectory contamination and cross-trajectory interference, which severely destabilizes policy updates. To overcome these bottlenecks, we introduce BIRD-RL, a novel four-stage agentic RL curriculum that alternately enhances foundational agent reasoning and trajectory-adaptive on-policy exploration to circumvent the instability inherently associated with direct end-to-end learning. Crucially, to support this curriculum, BIRD-RL is driven by a Trajectory-scoped Persistent Execution Infrastructure, which enables safe database replica reuse while preserving trajectory isolation. This design improves spatial RL efficiency by 12.8x, successfully enabling scalable agentic RL for complex data-centric tasks without needs of Kubernettes. Empirical evaluations on the BIRD-CRITIC-SQLite benchmark demonstrate that our 7B and 14B models achieve Success Rates of 44.40% and 48.00%, respectively, showing competitive performance against strong proprietary model-based agents such as Claude-Opus-4.6 Agent (48.20%). Furthermore, BIRD-RL successfully trains unified SLMs capable of both advanced SQL debugging and generation, highlighting its broader potential for multi-objective optimization across heterogeneous, real-world data-centric tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29426
PDF
-
LLMSeg: Training-Free Text-Prompted Image Segmentation with B-Splines
2026-10-21 08:35:16Abdullah HamdiAbstract
Promptable segmentation models such as SAM 3 achieve remarkable open-vocabulary performance but require dedicated architectures, large-scale annotation engines, and task-specific training. We ask a different question: can a general-purpose multimodal large language model (MLLM) segment images directly, without an added segmentation head, mask decoder, or training? We present LLMSeg, a training-free pipeline that turns a vision-capable LLM into a text-prompted image segmenter. LLMSeg decomposes segmentation into three LLM-driven stages. First, detect localizes all instances matching a free-form text query and describes their coarse morphology. Next, segment traces each instance using closed periodic B-splines confined to its detected box. These splines define an outer boundary and optional hole loops. Finally, refine iteratively proposes focus boxes around residual errors and retraces confined splines. The process ends when the mask stabilizes or the pass budget is exhausted. The output is a binary mask with inspectable intermediate geometry. We evaluate LLMSeg across ten referring and reasoning segmentation splits (ReasonSeg and RefCOCO/+/g). We also evaluate open-vocabulary instance segmentation (LVIS, COCO, and SA-Co), semantic segmentation, and zero-shot detection (ODinW13 and RF-100VL). With an off-the-shelf Gemini-3.7-Flash backend, LLMSeg is competitive with and often surpasses the specialist SAM 3 model on identical items. It reaches 71.5 cIoU on ReasonSeg val without additional segmentation-specific training. Backend model updates can improve accuracy without retraining the pipeline. The code is available at this https URL .
Abstract (translated)
URL
https://arxiv.org/abs/2610.29409
PDF
-
Right On Q: Recovering Reasoning under Quantization via Self-Generated Trajectories
2026-10-21 08:03:24Boyuan Zheng, Jing Xiong, Chenwei Wu, Hui Shen, Zesen Zhao, Hei Ting Una Chan, Xueshen Liu, Zhen Zhang, Liyue Shen, Zechun Liu, Z. Morley MaoAbstract
Quantization-aware training (QAT) aims to recover reasoning capabilities in low-bit large language models (LLMs). Existing recipes typically supervise the quantized student with the original full-precision model on fixed response sequences. However, the resulting low-bit quantized models still struggle on long-horizon reasoning tasks and often fail to terminate within a token budget that suffices for the full-precision model. We identify two challenges in closing this gap. First, fixed-response supervision leaves a mismatch between training prefixes and those encountered during student generation, and this trajectory divergence grows as responses lengthen. Second, on its own prefixes, the student underpredicts some next-token candidates and overpredicts others, and this distributional misalignment becomes more frequent at high-entropy prefixes. We present OnQ, a low-bit recovery pipeline that combines on-policy distillation and candidate-wise Dynamic KL. Starting from block-wise calibration, the quantized student generates responses and receives full-precision teacher supervision on its own prefixes. Dynamic KL weights forward and reverse KL by each candidate's teacher-student probability ratio to balance correction of under- and overprediction. Across Qwen3-1.7B, 4B, and 8B at 2- and 3-bit precision, OnQ achieves the highest nine-benchmark average, with the largest gains on long-horizon reasoning benchmarks. At 2 bits, it outperforms every baseline on every benchmark at every model size, improving the Qwen3-8B average by 10.26 points over the strongest baseline. It also produces shorter responses with fewer truncations and repetitive errors.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29378
PDF
-
Reading the Mask: Descriptor-Grounded BI-RADS Rationale Generation for Breast MRI Segmentation
2026-10-21 07:50:20Hrishikesh Singh, Aarti Sharma, Divyanshu Kumar, Millie PantAbstract
Deep learning based models accurately delineate tumours on dynamic contrast-enhanced MRI (DCE-MRI), but segmentation masks alone do not provide the rationales or reasoning used by radiologists to assess the lesion suspicion. This limits the explainability as well the clinical integration. We investigate whether clinically relevant descriptors can instead be derived from the segmentation masks and their dynamic contrast-enhancement series. We have used Breast Imaging Reporting and Data System (BI-RADS) guidelines as a structured framework. Using these guidelines we designed a BI-RADS decision tree with eight quantitative measurements using radiomics, together with the lesion size and component count, that determine the descriptors at eight nodes. Using these descriptors we generate natural-language rationale using a large language model. The model was constrained to express only the derived records from the decision tree. After applying this framework to the expert and automatic masks of each of the 1,506 cases in the MAMA-MIA dataset, we generated a reproducible corpus of 3,012 BI-RADS grounded rationales. We obtained a faithfulness score of 99.98% and a counterfactual accuracy of 99.3%, indicating that the generated rationales faithfully verbalises the measured descriptors and correctly reflects changes in the underlying record. Agreement between the descriptors from expert and automatic masks falls from 0.832 to 0.549 as mask quality falls from Good to Missed. The generated rationales are intended as draft representations for radiologist review and final sign-off rather than autonomous clinical reports.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29358
PDF
-
Decodability Is Not Controllability: Local and Anisotropic Speaker Control in Continuous-Latent TTS
2026-10-21 07:38:25Yuanzhong Zheng, Wei Huang, Ming Ma, Kailun Jin, Youhai JiangAbstract
High linear decodability does not entail causal controllability in continuous-latent text-to-speech (TTS). In VoxCPM2, position-matched probes and equal-norm interventions across 30 speakers and 60 directed dyads expose this separation: reference-audio states at the final residual acoustic language model layer (RALM L7) retain 0.970 probe accuracy, yet aligned edits yield 0/60 target-dominant dyads. Edits at early reference-audio positions produce the opposite pattern. At L0, pair-aligned updates increase target similarity by 0.326 and achieve target dominance in all 60 dyads, against at most 1/60 under any equal-norm direction or position control. Independent speaker encoders, retrieval against held-out speaker prototypes, and double-blind listening confirm that the shifted voice is heard as the intended target. Cross-architecture interventions in this http URL distinguish backbone leverage from direct acoustic control: backbone edits displace the source without target attraction, while x-vector ablation reduces source similarity from 0.849 to 0.006. These findings identify two requirements for training-free speaker control: geometric specificity and a computational route that carries the intended speaker condition into acoustic synthesis.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29345
PDF
-
Uncensored Is Not Unbiased: Abliteration Removes Refusals but Not Script-Dependent Bias About Taiwan
Abstract
"Abliterated" ("uncensored") builds of open-weight language models remove the activation direction that mediates refusal, and are popular partly on the assumption that a model that no longer refuses is also free of political censorship. We test this on two matched pairs of 2-bit models: NVIDIA's Nemotron 3.5 Bolt 74B-A9B and Qwen3.8-Flash-Next, each against a community abliteration. Each model judged 22 statement pairs (on the Republic of China (Taiwan), 1989 Tiananmen, control countries) in Traditional Chinese, Simplified Chinese (a character-level conversion) and English, and answered 27 open questions per script. Script mattered far more than abliteration: the original Bolt scored +0.58 on Index S (-1: People's Republic of China (PRC) position; +1: ROC position) in Traditional Chinese, +0.57 in English and +0.05 in Simplified Chinese, and affirmed "Taiwan is part of the PRC" with probability 0.91 in Simplified but 0.06 in Traditional characters; abliteration changed its scores by at most 0.06, mostly by raising "yes" on both sides. In Qwen, abliteration removed every refusal on 1989 and raised scores by 0.21 to 0.34, yet the abliterated Qwen still affirmed that Taiwan is part of the PRC in all three scripts. Naming the state the Republic of China (Taiwan) raised every model's score by 0.10 to 0.33 but left the script gap. Control pairs showed no script gap; each abliteration changed only its declared tensors. In 740,000 web documents, Simplified text on Taiwan contained PRC-framing terms 3.7 times as often as Traditional text. A script direction could not be steered away; a contrastive stance direction raised scores in every script, and its negation lowered them. A four-sentence system prompt raised the abliterated Qwen to +0.93 to +0.97. Keyword scoring rules failed the models closest to the ROC position mostly for attributing or denying the PRC claim. Uncensored is not unbiased.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29342
PDF
-
From Prompts to Posteriors: Can LLM Agents Perform Bayesian Inference in Context?
2026-10-21 07:29:24Linus Filbry, Luckeciano C. Melo, Yarin GalAbstract
Large language models (LLMs) are increasingly applied across the scientific discovery pipeline, from synthesizing literature to generating hypotheses to informing experimental design. A unifying principle behind these tasks is the need to continuously update beliefs as evidence accumulates, for which Bayesian inference provides a natural framework. However, existing work assumes that LLMs cannot carry out principled Bayesian belief updating in context, instead relying on simplifications and heuristics. To challenge this assumption, we present a principled evaluation of Bayesian reasoning capabilities in LLMs, analyzing both the model's internal consistency as well as its ability to recover an externally determined posterior. Specifically, we build an evaluation framework spanning qualitative epistemological examples to probe probabilistic coherence, quantitative benchmarks of Bayesian inference capabilities, and analyses of the impact of non-Bayesian updates on downstream task performance. Across different scenarios, we find that LLMs generally violate Bayes' rule when updating their beliefs in context. However, when equipped with modern reasoning abilities and code execution capabilities, their performance drastically improves, enabling even the close approximation of analytically intractable posteriors. Our evaluation shows that current LLMs with access to appropriate tools could already be used to make complex belief updates in scientific discovery, enabling rapid acceleration of the efficiency and scale of AI-powered scientific discovery.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29331
PDF
-
Plan First, Act Later: Streaming Reasoning for Executable Memory in Robot Control
2026-10-21 07:27:58Qinghang Liu, Conglin Wang, Zongsheng Li, Boyong He, Han Gao, Pengcheng Wang, Tao Sun, Lige Liu, Chang LiuAbstract
Long-horizon robot manipulation requires a robot to remember task-relevant information beyond its short observation window. Existing memory-augmented policies typically retrieve relevant information from history and reason over it at execution time, placing a heavy reasoning burden on the high-level model. To address this challenge, we propose PAL (\textbf{P}lan first, \textbf{A}ct \textbf{L}ater), a hierarchical memory framework for vision-language robot manipulation. It maintains a short-term visual buffer, a semantic-event memory, and an executable \texttt{subtask\_list}. A high-level vision-language model (VLM) converts streaming visual observations into the \texttt{subtask\_list}. At execution time, the VLM selects the currently relevant subtask from the \texttt{subtask\_list} and passes it to a low-level vision-language-action (VLA) policy. To teach the model what to write to memory and when to update it, we introduce a two-stage data generation procedure that first expands sparse update supervision and then uses self-labeling with task-grounded verification to generate training data that better match the state distribution encountered during execution. We evaluate our approach on memory-dependent benchmarks and real-world manipulation tasks. On RoboMME, PAL achieves a \textbf{56.05\%} success rate, outperforming the strongest baseline, FrameSamp-Modul, at 44.51\%. In real-world experiments across four manipulation tasks, \pal{} reaches \textbf{60.83\%} overall success, surpassing MemER (45.00\%) and FrameSamp-Modul (28.33\%). These results show that proactively writing future-relevant subtasks into hierarchical memory improves long-horizon visual reasoning and robot control. Code and videos can be found on our website: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.29329
PDF
-
Asking What a Director Sees Can Change What Vision-Language Models Choose
2026-10-21 07:21:08Jiaxuan Guo, Xingyu Ren, Kejia Zhang, Jingxin Yang, Youran Sun, Haizhao YangAbstract
Vision-language models are now routinely tested on theory of mind: whether they can work out what another person sees, knows, and means. A common design asks the model to report that perspective and then to act for the person, and reads a correct report beside a wrong action as a failure to use the perspective. That reading rests on a quiet assumption: that asking for the report leaves the action unchanged. We show that it does not. In a rendered Director Task, a partition hides the largest block from a director who asks for "the largest block". Asking first which blocks the director can see raises the share of replies that choose this hidden block by 12 to 88 percentage points on six GPT models and by 17 on Qwen3.8-27B, even though the models mostly report correctly which blocks the director sees. Reasoning traces and prompt rewordings suggest why: the question shifts how the request is read, from the director's view toward the model's own. When the prompt states what the director knows and refers to, fewer than 1% of replies pair a correct report with a hidden-block choice. Any evaluation that elicits a perspective report before the action it scores needs a baseline without the question, and we deliver one: paired prompts that differ only in that question, a design any perspective-taking benchmark can adopt.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29318
PDF
-
Alpha: Systematic Fault Injection and Robustness Evaluation for LLM-Driven Embodied Manipulation
Abstract
Large language models (LLMs) are increasingly integrated into embodied robotic planning, yet standard benchmarks primarily evaluate end-to-end task completion in nominal environments, obscuring failure origins across the execution stack. We present Alpha, a systematic fault-injection and robustness-evaluation framework for LLM-driven manipulation in simulation. Alpha injects calibrated faults across perception (sensor lag), actuation (grip slip), kinematics (unreachable targets), and planning (planner-output corruption), alongside concurrent conditions. To separate physical vulnerabilities from planner sensitivities, we introduce a matched methodology pairing an LLM agent (Llama-3 via Ollama) against a deterministic Scripted Oracle planner across a balanced 150-episode matrix under identical seeds, geometries, and safety criteria. Across the matrix, the Oracle achieves 64.0% (48/75) success while the LLM achieves 44.0% (33/75) (paired McNemar p = 3.01 x 10^-4). Crucially, excluding planner-output corruption, performance across all physical and kinematic conditions is identical: 55.0% (33/60) for both planners with zero discordant pairs (McNemar p = 1.00). This demonstrates that physical actuator slip, kinematic reach boundaries, and perceptual staleness affect deterministic controllers and LLM planners equivalently. The performance gap is driven entirely by planner-output corruption (100% Oracle vs. 0% LLM), isolating the prompt-to-JSON serialization and validation interface rather than a cognitive reasoning deficit. We also audit 169 historical pilot attempts for transparency.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30197
PDF
-
LLM-Based Exogenous World Modeling for Dynamic Scheduling
2026-10-21 07:01:25Ni Zhang, Haofeng Yuan, Jianan Zhou, Shiji Song, Zhiguang CaoAbstract
Dynamic scheduling requires dispatching online arrivals while anticipating future jobs, as poor decisions can compound downstream costs under nonlinear tardiness objectives. We show that a pretrained large language model (LLM) can serve as the exogenous component of a structured world model, forecasting uncertain future job arrivals. Specifically, we exploit an afterstate decomposition, which separates known deterministic shop transitions from uncertain exogenous job arrivals, enabling the LLM to forecast only the latter through a compact belief representation. Building on this, we develop Afterstate World-Model Planning (AWMP), which combines an evolved, belief-conditioned dispatching strategy with short scenario-based rollouts. A surprise-triggered refresh mechanism queries the LLM only when newly observed arrivals deviate sufficiently from previous forecast, and online residual self-calibration further conditions the LLM on its recent forecast errors to handle nonstationary arrivals. Across diverse arrival distributions, AWMP reduces weighted tardiness by 6-21% on dynamic job-shop and 22-46% on flexible job-shop benchmarks, relative to the best of eight classical priority-dispatching rules.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29308
PDF
-
ExpEvolve: Structuring and Reusing Experience for Open-Ended Discovery
2026-10-21 07:00:23Yongqiang Chen, Chenxi Liu, Deyu Zou, Zhenhao Chen, Tongliang Liu, Bo Han, Kun ZhangAbstract
Evolutionary coding agents such as AlphaEvolve are one of the notable successes in using Large Language Models (LLMs) to build AI Scientists. These agents tackle open-ended scientific problems by iteratively improving and evolving programs, leveraging the prior knowledge and reasoning capabilities of LLMs. Despite their success, mutations over programs are mostly low-level upon surface heuristics, without a persistent unit of control. In contrast, science compounds through ideas. In this work, we propose ExpEvolve, which injects idea-level extraction and control into evolution: it discovers and leverages the high-level factors from programs that explain scoring differences among different programs. We characterize theoretical conditions on when discovery with idea factors improves sample efficiency. We also build a practical framework to show that the extracted high-level idea factors are helpful for guiding exploitation from experience and exploration when discovery comes to a plateau. Across 11 open-ended mathematical tasks, ExpEvolve significantly improves the final performance and sample efficiency up to 5.0x. The improvements also transfer to 4 other LLMs and to 6 machine learning tasks from MLS-Bench. Analysis shows that ExpEvolve can discover idea-level factors to provide better exploration and exploitation for evolutionary discovery.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29303
PDF
-
Compressing Onboard Language Models for Space Robots under a Behavioral-Fidelity Requirement
2026-10-21 06:59:07Jia Deng, Zizhe Wang, Xin Lu, Yi Huang, Zehao JiaAbstract
Language models are being explored as the interface through which space robots interpret operator commands and select among verified skills. Running them onboard calls for low-bit compression, and flight use adds requirements that generic post-training quantization leaves implicit: a fixed memory budget and storage format, a static artifact under configuration control, and behavior close to the full-precision model validated on the ground. We treat compression as a ground-segment step and study QSelect, which composes one static low-bit model from compatible AWQ, GPTQ, and AutoRound checkpoints by assigning each linear module to one quantizer family and scoring every substitution at the final next-token distribution. The default fidelity mode needs no labels or tuned thresholds and accepts a substitution only when the forward KL divergence from the BF16 reference decreases; an optional target-adapted mode adds answer likelihood on a small labeled set. On DeepSeek-R1-Distill-Qwen-7B, Llama-3.1-8B, and Llama-3.2-3B at 3 and 4 bits, composed models reach lower mean held-out perplexity than every single-family model in all six settings, and the fidelity mode lowers perplexity and KL further. In a matched scaffold, final-output scoring gives lower held-out KL than representation-similarity scoring in all 12 paired runs. With labels from training or development splits of the evaluated task families, the target-adapted mode raises 3-bit pooled accuracy on the two Llama models by 4.3 and 4.5 points over the best single-family model. At 3 bits, unquantized embedding and output layers make up 43-46% of the packed artifact. All evaluations use general-domain benchmarks on a ground GPU; space-domain tasks and flight hardware are left to future work.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29302
PDF
-
Same Probabilities, Different Advice: Gain-Loss Framing Effects in High-Stakes LLM Decision Support
2026-10-21 06:46:33Yuyao Zhang, Yakun Cui, Han Zhu, Jinsong Lu, Yike Guo, Sirui HanAbstract
Large language models (LLMs) are increasingly used to support consequential decisions, yet their reliability under equivalent representations of uncertainty remains unclear. Existing work has documented prompt sensitivity, cognitive biases, and confidence miscalibration, but it is still unknown whether high-stakes advice changes when the same probability is expressed through complementary gain and loss frames. We introduce FrameBench-HPS, a controlled benchmark of 2,880 prompts derived from 240 high-stakes scenarios across healthcare, finance, education/career, and legal/consumer-rights domains. Across seven LLMs, gain-loss framing changes the recommendation category in 32.4% of complete groups, with most changes involving uncertainty rather than direct reversals. Human evaluation confirms corresponding shifts in perceived risk and directiveness. Robustness experiments show that framing effects persist under conversational paraphrases and exceed ordinary repeated-generation variability. Probability canonicalization substantially reduces framing sensitivity, identifying probability representation as a practical target for improving consistency in high-stakes LLM decision support.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29294
PDF
-
Computer Vision and Knowledge-Grounded Language Models for Intelligent Beekeeper Assistance
2026-10-21 06:36:55Asaduz Zaman, Mohaimenul Azam Khan Raiaan, Malika Nisal RatnayakeAbstract
Managed and wild bees sustain roughly a third of the global food products, yet colonies face mounting stress from climate-change including extreme weather patterns, altered parasite and disease patterns. Timely assessment of colony condition such as understanding colony strength, brood area, honey and pollen storage, comb usage, is essential to adaptive management. However, manual hive frame inspection is slow, costly, and quality depends on scarce expertise. We present an artificial targeted intelligent system framework focusing on edge deployable pipeline that turns hive frame photographs into interpretable and action-oriented inspection report. The pipeline includes an object detector to detect biologically-meaningful comb-cell classes; detections are distilled into structured fact schema; facts are converted into constrained factual text; the text triggers a grounded knowledge-base for interpretations and recommendations. Knowledge-base entries are required to be triggerable by evidence expressed in the fact schema. We built our system based on open source FAIRHiveFrames-1K dataset and favour reproducible, edge-deployable models over large general purpose frontier models. We additionally release FAIRHiveText-1K dataset that turns FAIRHiveFrames-1K dataset into interchangeable facts and text interpretation of the facts, making it easy for further LLM-based research into this area. We also propose a pathway to develop the grounded knowledge-base that will be helpful in providing interpretable and actionable reports from the facts.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29283
PDF
-
Flat-RLVR: Sharpness-Aware Reinforcement Learning with Verifiable Rewards for Reasoning Capacity in LLMs
2026-10-21 06:35:19Yeonbum Kim, Hyun Kyu Lee, Sung Whan YoonAbstract
Reinforcement learning with verifiable rewards (RLVR), particularly through group-relative policy optimization (GRPO), has emerged as a central paradigm for the post-training of large language models (LLMs). While RLVR is highly effective at enhancing the reasoning capabilities of LLMs, it commonly induces \textit{distribution sharpening}, whereby the policy distribution becomes excessively peaked and rare-yet-correct trajectories are progressively suppressed, thus diminishing sample diversity. To this end, we promote flat-reward optima in RLVR to smooth the policy distribution. We theoretically link flat reward RLVR to reduced distribution sharpening, and empirically show improved sample diversity while maintaining single-attempt accuracy. Our method, called \texttt{Flat}-RLVR, searches for flat reward regions via sharpness-aware minimization during LLM post-training, combining this with GRPO variants to achieve better trade-offs between sampling efficiency and reasoning capacity.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29281
PDF
-
ARCANA: A Benchmark for Abstraction and Analogical Reasoning
2026-10-21 06:18:46Ruqi Chen, Zilei Wang, Wei WangAbstract
Analogical reasoning is widely considered a fundamental component of human intelligence. Few benchmarks are designed to explicitly test far analogical reasoning - reasoning between domains that are superficially dissimilar yet share deep relational or structural similarities. In this work, we introduce an analogical reasoning benchmark called Abstraction and Reasoning Corpus of ANAlogy (ARCANA). The tasks of ARCANA require performing far analogical mappings from real-world objects or physical phenomena to abstract grid-based transformations, different from the well-known ARC-AGI benchmark which primarily emphasized innate human priors. We validate ARCANA through a human study, demonstrating that humans solve these tasks reliably. In contrast, evaluations across a range of large language models show a distinct performance gap compared to humans, revealing systematic failure modes that remain challenging for current models. To address this gap, we propose ConceptSIR, a reasoning framework based on concept slippage, introspection, and reflection. Experiments demonstrate that ConceptSIR enables standard LLMs to solve several challenging ARCANA tasks that were unsolved by direct prompting in our evaluation. Overall, our results suggest that far analogical reasoning remains an open problem for contemporary LLMs. All the materials are available at this http URL. The complete codebase, including benchmark, and ConceptSIR, is publicly available at this https URL. All ARCANA data and task instances can be accessed and explored at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29271
PDF
-
Heard But Not Said: Prompting Understates Temporal Information of Musical Moments in Frozen Large Audio-Language Models
Abstract
Large audio-language models (LALMs) understand music, yet struggle to locate a described moment in time. We ask whether this reflects missing temporal information or a bottleneck in how the time span is extracted. Using a paired-query probe built from Slakh2100 stem activations, we attach a state query (e.g., ``only piano and bass are playing'') and a change query (e.g., ``the guitar drops out, leaving piano and bass'') to the same span and vary how much preceding audio the model is shown. Across frozen LALMs, three ways of extracting the span ask different questions: full-track prompting asking whether the model can say when; windowed yes/no scoring asking whether it verifies the moment when handed the segment, and whether it needs what came before; attention asking whether it can find the moment on its own. State queries score above baseline without context, while change queries improve with preceding context in most short-segment models; 30s padded encoders show no comparable recovery. Prompting and attention expose almost none of this. Prompting failures thus understate what LALMs can verify, though not what they can locate, by an amount set by query type and encoder.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29256
PDF
-
Hierarchical Masked Diffusion Language Models
2026-10-21 06:06:01Parul Gupta, Satwik, Avni Aggarwal, Bhavesh Gurnani, Arnab Kumar Mondal, Dinesh Khandelwal, Dinesh Garg, Parag SinglaAbstract
Masked diffusion language models have emerged as an alternative to autoregressive models, enabling parallel token generation. However, standard masked diffusion language models use a shared noise schedule across positions without explicitly encoding which positions should be resolved first. Many tasks exhibit a natural priority among positions: document titles and section headings can guide the generation of body text, while Sudoku cells determined directly from the initial clues can help resolve more difficult cells. To incorporate this notion of priority (or hierarchy) among token positions, we introduce the Hierarchical Masked Diffusion Language Model (HMDLM), which augments each token position with a hierarchy level available as part of the training data, and requires zero or minimal additional annotations. It factors generation into two passes: the first predicts a hierarchy distribution over masked positions, the second predicts token distributions conditioned on it. Giving each hierarchy level its own noise schedule induces a soft, controllable preference for revealing higher-priority positions earlier while keeping decoding fully parallel. Our formulation is mathematically grounded, and we re-derive the forward and reverse processes and the NELBO for this factorization. Our evaluation shows that for the task of language generation on LM1B, HMDLM reduces perplexity by more than 1.5 points over Anchored Diffusion Language Model (ADLM), a recently proposed strong diffusion baseline; on WikiHow, we see a drop in perplexity of up to 4.4 pts on ADLM and up to 7 pts against MDLM. An LLM judge on WikiHow prefers HMDLM generations (the two variants combined) more than $2\times$ as often as the baseline. On the task of solving Sudoku puzzles requiring multi-step reasoning, it improves solution accuracy by up to $8.6\%$ absolute points over existing diffusion baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29255
PDF
-
Failure Modes of Vision-Language Models for Human Activity Understanding in Thermal HRI
2026-10-21 05:49:05Han Hu, He ChenAbstract
Thermal cameras enable robots to perceive human activities in darkness while revealing less personally identifiable appearance information and fewer details of the surrounding environment than RGB cameras. As vision-language models (VLMs), originally developed for RGB imagery, are increasingly integrated into embodied systems for scene understanding and high-level planning, thermal sensing offers a promising way to extend activity perception to low-light and privacy-sensitive settings. However, the failure modes of VLMs in thermal human activity understanding for human-robot interaction (HRI) remain underexplored. We fill this gap by evaluating seven locally deployable VLMs, including models developed for physical AI and robotics, on two person-centric thermal datasets. Our diagnostic tasks examine fine-grained action recognition, activity timeline reconstruction, and the ability to withhold unsupported answers. Paired RGB recordings serve as a performance reference, while controlled variations in frame sampling, thermal input presentation, and reasoning are used to assess whether the observed errors can be mitigated. Our results show that models can recognize broad activity categories but struggle with fine-grained actions such as coughing, sneezing, and yawning. They also produce inaccurate event timelines and hallucinate actions involving objects. Simple thermal input preprocessing and additional reasoning offer only limited, model-dependent gains. Abstention behavior also varies across models even under explicit instructions, limiting their reliability for practical robotic use.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29244
PDF
